#include "i915_gem_clflush.h"
#include "i915_gem_context.h"
#include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
#include "i915_trace.h"
#include "i915_user_extensions.h"
-#include "i915_memcpy.h"
struct eb_vma {
struct i915_vma *vma;
int err;
struct intel_engine_cs *engine = eb->engine;
+ /* If we need to copy for the cmdparser, we will stall anyway */
+ if (eb_use_cmdparser(eb))
+ return ERR_PTR(-EWOULDBLOCK);
+
if (!reloc_can_use_engine(engine)) {
engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
if (!engine)
return vma;
}
-struct eb_parse_work {
- struct dma_fence_work base;
- struct intel_engine_cs *engine;
- struct i915_vma *batch;
- struct i915_vma *shadow;
- struct i915_vma *trampoline;
- unsigned long batch_offset;
- unsigned long batch_length;
- unsigned long *jump_whitelist;
- const void *batch_map;
- void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
- struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
- int ret;
- bool cookie;
-
- cookie = dma_fence_begin_signalling();
- ret = intel_engine_cmd_parser(pw->engine,
- pw->batch,
- pw->batch_offset,
- pw->batch_length,
- pw->shadow,
- pw->jump_whitelist,
- pw->shadow_map,
- pw->batch_map);
- dma_fence_end_signalling(cookie);
-
- return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
- struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
- if (!IS_ERR_OR_NULL(pw->jump_whitelist))
- kfree(pw->jump_whitelist);
-
- if (pw->batch_map)
- i915_gem_object_unpin_map(pw->batch->obj);
- else
- i915_gem_object_unpin_pages(pw->batch->obj);
-
- i915_gem_object_unpin_map(pw->shadow->obj);
-
- if (pw->trampoline)
- i915_active_release(&pw->trampoline->active);
- i915_active_release(&pw->shadow->active);
- i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
- .name = "eb_parse",
- .work = __eb_parse,
- .release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
- struct intel_timeline *tl,
- struct dma_fence *fence)
-{
- struct intel_gt_buffer_pool_node *node = vma->private;
-
- return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
- int err;
-
- mutex_lock(&tl->mutex);
-
- err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
- if (err)
- goto unlock;
-
- if (pw->trampoline) {
- err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
- if (err)
- goto unlock;
- }
-
-unlock:
- mutex_unlock(&tl->mutex);
- return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
- struct i915_vma *shadow,
- struct i915_vma *trampoline)
-{
- struct eb_parse_work *pw;
- struct drm_i915_gem_object *batch = eb->batch->vma->obj;
- bool needs_clflush;
- int err;
-
- GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
- GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
- pw = kzalloc(sizeof(*pw), GFP_KERNEL);
- if (!pw)
- return -ENOMEM;
-
- err = i915_active_acquire(&eb->batch->vma->active);
- if (err)
- goto err_free;
-
- err = i915_active_acquire(&shadow->active);
- if (err)
- goto err_batch;
-
- if (trampoline) {
- err = i915_active_acquire(&trampoline->active);
- if (err)
- goto err_shadow;
- }
-
- pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
- if (IS_ERR(pw->shadow_map)) {
- err = PTR_ERR(pw->shadow_map);
- goto err_trampoline;
- }
-
- needs_clflush =
- !(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
- pw->batch_map = ERR_PTR(-ENODEV);
- if (needs_clflush && i915_has_memcpy_from_wc())
- pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
- if (IS_ERR(pw->batch_map)) {
- err = i915_gem_object_pin_pages(batch);
- if (err)
- goto err_unmap_shadow;
- pw->batch_map = NULL;
- }
-
- pw->jump_whitelist =
- intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
- trampoline);
- if (IS_ERR(pw->jump_whitelist)) {
- err = PTR_ERR(pw->jump_whitelist);
- goto err_unmap_batch;
- }
-
- dma_fence_work_init(&pw->base, &eb_parse_ops);
-
- pw->engine = eb->engine;
- pw->batch = eb->batch->vma;
- pw->batch_offset = eb->batch_start_offset;
- pw->batch_length = eb->batch_len;
- pw->shadow = shadow;
- pw->trampoline = trampoline;
-
- /* Mark active refs early for this worker, in case we get interrupted */
- err = parser_mark_active(pw, eb->context->timeline);
- if (err)
- goto err_commit;
-
- err = dma_resv_reserve_shared(pw->batch->resv, 1);
- if (err)
- goto err_commit;
-
- err = dma_resv_reserve_shared(shadow->resv, 1);
- if (err)
- goto err_commit;
-
- /* Wait for all writes (and relocs) into the batch to complete */
- err = i915_sw_fence_await_reservation(&pw->base.chain,
- pw->batch->resv, NULL, false,
- 0, I915_FENCE_GFP);
- if (err < 0)
- goto err_commit;
-
- /* Keep the batch alive and unwritten as we parse */
- dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
- /* Force execution to wait for completion of the parser */
- dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
- dma_fence_work_commit_imm(&pw->base);
- return 0;
-
-err_commit:
- i915_sw_fence_set_error_once(&pw->base.chain, err);
- dma_fence_work_commit_imm(&pw->base);
- return err;
-
-err_unmap_batch:
- if (pw->batch_map)
- i915_gem_object_unpin_map(batch);
- else
- i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
- i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
- if (trampoline)
- i915_active_release(&trampoline->active);
-err_shadow:
- i915_active_release(&shadow->active);
-err_batch:
- i915_active_release(&eb->batch->vma->active);
-err_free:
- kfree(pw);
- return err;
-}
-
static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
{
/*
goto err_trampoline;
}
- err = eb_parse_pipeline(eb, shadow, trampoline);
+ err = dma_resv_reserve_shared(shadow->resv, 1);
+ if (err)
+ goto err_trampoline;
+
+ err = intel_engine_cmd_parser(eb->engine,
+ eb->batch->vma,
+ eb->batch_start_offset,
+ eb->batch_len,
+ shadow, trampoline);
if (err)
goto err_unpin_batch;
intel_gt_pm_get(&eb.i915->gt);
for_each_uabi_engine(eb.engine, eb.i915) {
+ if (intel_engine_requires_cmd_parser(eb.engine) ||
+ intel_engine_using_cmd_parser(eb.engine))
+ continue;
+
reloc_cache_init(&eb.reloc_cache, eb.i915);
memset(map, POISON_INUSE, 4096);
if (drm_WARN_ON(&i915->drm, !engine))
return -EINVAL;
+ /*
+ * Due to d3_entered is used to indicate skipping PPGTT invalidation on
+ * vGPU reset, it's set on D0->D3 on PCI config write, and cleared after
+ * vGPU reset if in resuming.
+ * In S0ix exit, the device power state also transite from D3 to D0 as
+ * S3 resume, but no vGPU reset (triggered by QEMU devic model). After
+ * S0ix exit, all engines continue to work. However the d3_entered
+ * remains set which will break next vGPU reset logic (miss the expected
+ * PPGTT invalidation).
+ * Engines can only work in D0. Thus the 1st elsp write gives GVT a
+ * chance to clear d3_entered.
+ */
+ if (vgpu->d3_entered)
+ vgpu->d3_entered = false;
+
execlist = &vgpu->submission.execlist[engine->id];
execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data;
static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
struct drm_i915_gem_object *src_obj,
unsigned long offset, unsigned long length,
- void *dst, const void *src)
+ bool *needs_clflush_after)
{
- bool needs_clflush =
- !(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
- if (src) {
- GEM_BUG_ON(!needs_clflush);
- i915_unaligned_memcpy_from_wc(dst, src + offset, length);
- } else {
- struct scatterlist *sg;
+ unsigned int src_needs_clflush;
+ unsigned int dst_needs_clflush;
+ void *dst, *src;
+ int ret;
+
+ ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+ if (ret)
+ return ERR_PTR(ret);
+
+ dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+ i915_gem_object_finish_access(dst_obj);
+ if (IS_ERR(dst))
+ return dst;
+
+ ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+ if (ret) {
+ i915_gem_object_unpin_map(dst_obj);
+ return ERR_PTR(ret);
+ }
+
+ src = ERR_PTR(-ENODEV);
+ if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+ src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+ if (!IS_ERR(src)) {
+ i915_unaligned_memcpy_from_wc(dst,
+ src + offset,
+ length);
+ i915_gem_object_unpin_map(src_obj);
+ }
+ }
+ if (IS_ERR(src)) {
+ unsigned long x, n, remain;
void *ptr;
- unsigned int x, sg_ofs;
- unsigned long remain;
/*
* We can avoid clflushing partial cachelines before the write
* validate up to the end of the batch.
*/
remain = length;
- if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+ if (dst_needs_clflush & CLFLUSH_BEFORE)
remain = round_up(remain,
boot_cpu_data.x86_clflush_size);
ptr = dst;
x = offset_in_page(offset);
- sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
- while (remain) {
- unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
- for (; remain && sg_ofs < sg_max; sg_ofs++) {
- unsigned long len = min(remain, PAGE_SIZE - x);
- void *map;
-
- map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
- if (needs_clflush)
- drm_clflush_virt_range(map + x, len);
- memcpy(ptr, map + x, len);
- kunmap_atomic(map);
-
- ptr += len;
- remain -= len;
- x = 0;
- }
-
- sg_ofs = 0;
- sg = sg_next(sg);
+ for (n = offset >> PAGE_SHIFT; remain; n++) {
+ int len = min(remain, PAGE_SIZE - x);
+
+ src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+ if (src_needs_clflush)
+ drm_clflush_virt_range(src + x, len);
+ memcpy(ptr, src + x, len);
+ kunmap_atomic(src);
+
+ ptr += len;
+ remain -= len;
+ x = 0;
}
}
+ i915_gem_object_finish_access(src_obj);
+
memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
/* dst_obj is returned with vmap pinned */
+ *needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
return dst;
}
if (target_cmd_index == offset)
return 0;
+ if (IS_ERR(jump_whitelist))
+ return PTR_ERR(jump_whitelist);
+
if (!test_bit(target_cmd_index, jump_whitelist)) {
DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
jump_target);
return 0;
}
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
- bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
{
unsigned long *jmp;
- if (trampoline)
- return NULL;
-
/*
* We expect batch_length to be less than 256KiB for known users,
* i.e. we need at most an 8KiB bitmap allocation which should be
* @batch_offset: byte offset in the batch at which execution starts
* @batch_length: length of the commands in batch_obj
* @shadow: validated copy of the batch buffer in question
- * @jump_whitelist: buffer preallocated with intel_engine_cmd_parser_alloc_jump_whitelist()
- * @shadow_map: mapping to @shadow vma
- * @batch_map: mapping to @batch vma
+ * @trampoline: true if we need to trampoline into privileged execution
*
* Parses the specified batch buffer looking for privilege violations as
* described in the overview.
* Return: non-zero if the parser finds violations or otherwise fails; -EACCES
* if the batch appears legal but should use hardware parsing
*/
+
int intel_engine_cmd_parser(struct intel_engine_cs *engine,
struct i915_vma *batch,
unsigned long batch_offset,
unsigned long batch_length,
struct i915_vma *shadow,
- unsigned long *jump_whitelist,
- void *shadow_map,
- const void *batch_map)
+ bool trampoline)
{
u32 *cmd, *batch_end, offset = 0;
struct drm_i915_cmd_descriptor default_desc = noop_desc;
const struct drm_i915_cmd_descriptor *desc = &default_desc;
+ bool needs_clflush_after = false;
+ unsigned long *jump_whitelist;
u64 batch_addr, shadow_addr;
int ret = 0;
- bool trampoline = !jump_whitelist;
GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
batch->size));
GEM_BUG_ON(!batch_length);
- cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
- shadow_map, batch_map);
+ cmd = copy_batch(shadow->obj, batch->obj,
+ batch_offset, batch_length,
+ &needs_clflush_after);
+ if (IS_ERR(cmd)) {
+ DRM_DEBUG("CMD: Failed to copy batch\n");
+ return PTR_ERR(cmd);
+ }
+
+ jump_whitelist = NULL;
+ if (!trampoline)
+ /* Defer failure until attempted use */
+ jump_whitelist = alloc_whitelist(batch_length);
shadow_addr = gen8_canonical_addr(shadow->node.start);
batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
i915_gem_object_flush_map(shadow->obj);
+ if (!IS_ERR_OR_NULL(jump_whitelist))
+ kfree(jump_whitelist);
+ i915_gem_object_unpin_map(shadow->obj);
return ret;
}
int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
- bool trampoline);
-
int intel_engine_cmd_parser(struct intel_engine_cs *engine,
struct i915_vma *batch,
unsigned long batch_offset,
unsigned long batch_length,
struct i915_vma *shadow,
- unsigned long *jump_whitelist,
- void *shadow_map,
- const void *batch_map);
+ bool trampoline);
#define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
/* intel_device_info.c */
do {
fence = *child++;
- if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
- i915_sw_fence_set_error_once(&rq->submit, fence->error);
+ if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
continue;
- }
if (fence->context == rq->fence.context)
continue;
do {
fence = *child++;
- if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
- i915_sw_fence_set_error_once(&rq->submit, fence->error);
+ if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
continue;
- }
/*
* Requests on the same timeline are explicitly ordered, along
projects / linux-2.6-microblaze.git / commitdiff